Gemini-compatible API

generateContent against the Asale gateway, with the official google-genai SDKs. Change the base URL and the key; your request bodies stay as they are.

The Gemini dialect is what google-genai and gemini-cli speak. Unlike the other two it carries both the model and the streaming flag in the path, not the body — which is the one thing to get right when you wire it up by hand.

Read the overview first for keys, models, billing and errors; everything here is the Gemini-specific part.

Connection


Base URL	`https://gw.asale.ai`
Auth header	`x-goog-api-key: sk-asale-...`
SDK env var	`GEMINI_API_KEY`, `GOOGLE_GEMINI_BASE_URL`

The base URL stops at the host. Google's clients append /v1beta/models/… themselves, so giving them https://gw.asale.ai/v1beta produces a doubled path and a 404. GOOGLE_GEMINI_BASE_URL — what gemini-cli reads — is an origin too.

Authorization: Bearer and x-api-key are accepted as well. A key in the ?key= query string is not — put it in a header.

Endpoints

Method	Path	Notes
POST	`/v1beta/models/{model}:generateContent`	The main one.
POST	`/v1beta/models/{model}:streamGenerateContent`	Streaming — the method name is the flag.
GET	`/v1beta/models`	What the platform will match right now.

Call it

curl "https://gw.asale.ai/v1beta/models/gemini-2.5-pro:generateContent" \
  -H "x-goog-api-key: $ASALE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{"role": "user", "parts": [{"text": "Hello"}]}]
  }'

# pip install google-genai
import os
from google import genai
from google.genai import types

client = genai.Client(
    api_key=os.environ["ASALE_API_KEY"],
    http_options=types.HttpOptions(base_url="https://gw.asale.ai"),
)

resp = client.models.generate_content(
    model="gemini-2.5-pro",
    contents="Hello",
)
print(resp.text)

// npm i @google/genai
import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({
  apiKey: process.env.ASALE_API_KEY,
  httpOptions: { baseUrl: "https://gw.asale.ai" },
});

const resp = await ai.models.generateContent({
  model: "gemini-2.5-pro",
  contents: "Hello",
});
console.log(resp.text);

Streaming

Call :streamGenerateContent instead of :generateContent. The method suffix is what turns streaming on — there is no stream field in a Gemini body — and the response is server-sent events carrying GenerateContentResponse chunks.

for chunk in client.models.generate_content_stream(
    model="gemini-2.5-pro",
    contents="Count to five",
):
    print(chunk.text, end="", flush=True)

Generation config and tools

generationConfig — maxOutputTokens, temperature, topP, stopSequences and the thinking budget pass through.
systemInstruction passes through as the system prompt.
tools / functionCall / functionResponse pass through in their native shape.

Practical notes

The model in the path wins. If the path names one model and the body another, the path is what you addressed and the path is what is served.
Set maxOutputTokens. It is what the balance hold is computed from; the difference returns when the response finishes.
Model names are the platform's, not Google's SKU list. GET /v1beta/models is the list; the market shows which of them have supply.
Request bodies are capped at 16 MiB.